CINXE.COM

<?xml version="1.0" encoding="UTF-8" standalone="yes"?> <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"><html xmlns="http://www.w3.org/1999/xhtml"><head><meta http-equiv="Content-Type" content="text/html; charset=UTF-8" /><title>LKML: Miklos Szeredi: Re: [RCF] [PATCH] unprivileged mount/umount</title><link href="/css/message.css" rel="stylesheet" type="text/css" /><link href="/css/wrap.css" rel="alternate stylesheet" type="text/css" title="wrap" /><link href="/css/nowrap.css" rel="stylesheet" type="text/css" title="nowrap" /><link href="/favicon.ico" rel="shortcut icon" /><script src="/js/simple-calendar.js" type="text/javascript"></script><script src="/js/styleswitcher.js" type="text/javascript"></script><link rel="alternate" type="application/rss+xml" title="lkml.org : last 100 messages" href="/rss.php" /><link rel="alternate" type="application/rss+xml" title="lkml.org : last messages by Miklos Szeredi" href="/groupie.php?aid=1702" /><script> var _paq = window._paq = window._paq || []; /* tracker methods like "setCustomDimension" should be called before "trackPageView" */ _paq.push(["setDoNotTrack", true]); _paq.push(["disableCookies"]); _paq.push(['trackPageView']); _paq.push(['enableLinkTracking']); (function() { var u="//m.lkml.org/"; _paq.push(['setTrackerUrl', u+'matomo.php']); _paq.push(['setSiteId', '1']); var d=document, g=d.createElement('script'), s=d.getElementsByTagName('script')[0]; g.async=true; g.src=u+'matomo.js'; s.parentNode.insertBefore(g,s); })(); </script></head><body onload="es.jasper.simpleCalendar.init();" itemscope="itemscope" itemtype="http://schema.org/BlogPosting"><table border="0" cellpadding="0" cellspacing="0"><tr><td width="180" align="center"><a href="/"><img style="border:0;width:135px;height:32px" src="/images/toprowlk.gif" alt="lkml.org" /></a></td><td width="32">聽</td><td class="nb"><div><a class="nb" href="/lkml"> [lkml]</a> 聽 <a class="nb" href="/lkml/2005"> [2005]</a> 聽 <a class="nb" href="/lkml/2005/5"> [May]</a> 聽 <a class="nb" href="/lkml/2005/5/13"> [13]</a> 聽 <a class="nb" href="/lkml/last100"> [last100]</a> 聽 <a href="/rss.php"><img src="/images/rss-or.gif" border="0" alt="RSS Feed" /></a></div><div>Views: <a href="#" class="nowrap" onclick="setActiveStyleSheet('wrap');return false;">[wrap]</a><a href="#" class="wrap" onclick="setActiveStyleSheet('nowrap');return false;">[no wrap]</a> 聽 <a class="nb" href="/lkml/mheaders/2005/5/13/9" onclick="this.href='/lkml/headers'+'/2005/5/13/9';">[headers]</a>聽 <a href="/lkml/bounce/2005/5/13/9">[forward]</a>聽 </div></td><td width="32">聽</td></tr><tr><td valign="top"><div class="es-jasper-simpleCalendar" baseurl="/lkml/"></div><div class="threadlist">Messages in this thread</div><ul class="threadlist"><li class="root"><a href="/lkml/2005/5/3/64">First message in thread</a></li><li><a href="/lkml/2005/5/11/61">Miklos Szeredi</a><ul><li><a href="/lkml/2005/5/12/175">Bryan Henderson</a><ul><li class="origin"><a href="/lkml/2005/5/13/22">Miklos Szeredi</a><ul><li><a href="/lkml/2005/5/13/22">Jan Hudec</a><ul><li><a href="/lkml/2005/5/13/43">Miklos Szeredi</a></li></ul></li></ul></li></ul></li></ul></li></ul></td><td width="32" rowspan="2" class="c" valign="top"><img src="/images/icornerl.gif" width="32" height="32" alt="/" /></td><td class="c" rowspan="2" valign="top" style="padding-top: 1em"><table><tr><td><table><tr><td class="lp">Subject</td><td class="rp" itemprop="name">Re: [RCF] [PATCH] unprivileged mount/umount</td></tr><tr><td class="lp">From</td><td class="rp" itemprop="author">Miklos Szeredi <></td></tr><tr><td class="lp">Date</td><td class="rp" itemprop="datePublished">Fri, 13 May 2005 07:47:07 +0200</td></tr></table></td><td></td></tr></table><pre itemprop="articleBody">> > So if a user creates a private namespace, it should have the choice of:<br />> > <br />> > 1) Giving up all suid rights (i.e. all mounts are cloned and<br />> > propagated with nosuid)<br />> > <br />> > 2) Not giving up suid for cloned and propagated mounts, but having<br />> > extra limitations (suid/sgid programs cannot access unprivileged<br />> > "synthetic" mounts)<br />> <br />> (2) isn't realistic. There's no such thing as a suid program. Suid is a <br />> characteristic of a _file_. There's no way to know whether a given <br />> executing program is running with privileges that came from a suid file <br />> getting exec'ed. Bear in mind that that exec could be pretty remote -- <br />> done by a now-dead ancestor with three more execs in between.<br />> <br />> Many user space programs contain hacks to try to discern this information, <br />> and they often cause me headaches and I have to fix them. The usual hacks <br />> are euid==uid, euid==suid, and/or euid==0. It would be an order of <br />> magnitude worse for the kernel to contain such a hack.<br /><br />Guess what? It's already there. Look in ptrace_attach() in<br />kernel/ptrace.c<br /><br />You could argue about the usefulness of ptrace. The point is, that<br />suid/sgid programs _can_ be discerned, and ptrace _needs_ to discern<br />them.<br /><br />And for the same reason, user controlled filesystems also need to do<br />this check. See Documentation/filesystems/fuse.txt in -mm and later<br />discussion in this thread for more information.<br /><br />The strange thing is that people argue so vehemently about this on<br />theoretical grounds. But practice shows, that there's in fact no<br />problem with it whatsoever. This check (or something similar), has<br />been in FUSE for more than 3 years, used by thousands of people, and<br />it didn't cause _any_ problems, except various sub-mount related<br />things, which we are trying to solve properly, by making mount<br />unprivileged.<br /><br />Miklos<br />-<br />To unsubscribe from this list: send the line "unsubscribe linux-kernel" in<br />the body of a message to majordomo@vger.kernel.org<br />More majordomo info at <a href="http://vger.kernel.org/majordomo-info.html">http://vger.kernel.org/majordomo-info.html</a><br />Please read the FAQ at <a href="http://www.tux.org/lkml/">http://www.tux.org/lkml/</a><br /><br /></pre></td><td width="32" rowspan="2" class="c" valign="top"><img src="/images/icornerr.gif" width="32" height="32" alt="\" /></td></tr><tr><td align="right" valign="bottom"> 聽 </td></tr><tr><td align="right" valign="bottom">聽</td><td class="c" valign="bottom" style="padding-bottom: 0px"><img src="/images/bcornerl.gif" width="32" height="32" alt="\" /></td><td class="c">聽</td><td class="c" valign="bottom" style="padding-bottom: 0px"><img src="/images/bcornerr.gif" width="32" height="32" alt="/" /></td></tr><tr><td align="right" valign="top" colspan="2"> 聽 </td><td class="lm">Last update: 2005-05-13 07:51 聽聽 [from the cache]<br />漏2003-2020 <a href="http://blog.jasper.es/"><span itemprop="editor">Jasper Spaans</span></a>|hosted at <a href="https://www.digitalocean.com/?refcode=9a8e99d24cf9">Digital Ocean</a> and my Meterkast|<a href="http://blog.jasper.es/categories.html#lkml-ref">Read the blog</a></td><td>聽</td></tr></table><script language="javascript" src="/js/styleswitcher.js" type="text/javascript"></script></body></html>